Dynamic pricing – machine learning
[Client prefers to remain anonymous due to competitive scope of project]
BLOOM has created a dynamic pricing algorithm for a rapidly growing company in the online retail space that handles thousands of products per day. The client was looking for a way to boost its revenues by using its historical sales data to dynamically adjust product prices.
In order to do so, BLOOM built an algorithm that determines the optimal price of products offered on multiple online channels. To calculate the best price, the algorithm takes into account product characteristics, historical selling data, and competitor behaviour. This results in over 2,000 features that together determine selling probability and potential revenue.
Under the hood, BLOOM has set up two gradient boosting models that work complementary: a regression model and a classification model. These models serve as input for a custom pricing algorithm that optimises expected return, taking into account today’s and future selling probabilities. The model is self-learning – it is updated daily with the latest selling data, thus continuously improving its accuracy.
The algorithm is now used many times every day to determine the optimal offer price for each product on each channel, allowing to benefit from different market conditions.
Back to cases overview
We were all very impressed with BLOOM’s advanced data science and machine learning skills. In a relatively short period, they were able to deliver an algorithm that directly improved our bottom line, while at the same time coaching our colleagues to embed the necessary skills in our team. Also, they’re a business-savvy, pragmatic, and fun bunch of people to work with. I would highly recommended working with the BLOOM team on machine learning and data science projects!
– Managing Director of client